In the middle of the good period the phase was adjusted by 0.15 rad, because the phase monitor was measuring a mistuning (Fig. 1). On MICH error signal the effect was a further small improvement (Fig. 2). A larger improvement was visible in SPRB_B4_56MHz_I, out-of-loop monitor of frequency noise (Fig. 3). No change at all in SSFS in-loop signal (Fig. 4). The phase was still well tuned later, when MICH noise was largely higher; the possibility that a further phase tuning would have reduced the noise seems not realistic.

By tha way, the different behavior of SSFS_Err and B4_56MHz_I seems interesting to investigate.

In Fig. 5 SSFS error signal is superposed to SPRB_B4_56MHz_I in the two different conditions of noise on MICH. In both cases the out of loop signal is higher than SSFS signal starting from a few hundred Hertz. This happens not only in the floor, but also for the frequency noise peaks, and the level of the peaks changes when 'MICH noise' changes.

The fact that the peaks are 'frequency noise' is shown in Fig. 6, where instead of SSFS error signal the RFC error signal is plotted (input mode cleaner stand alone, MC locked on RFC). The peaks which go up and down in B4 are already present in IMC stand alone.

Finally, in Fig. 7 B4_56MHz I and Q are shown again in the two different conditions of noise. The common noise which goes up and down seems to be frequency noise.